This paper describes a new landmine detection algorithm starting from high resolution #D ground penetrating radar (DPR) images. The algorithm consists of two procedures, object detection and object classification; both strongly depend on the properties of 3 GPR images. The algorithm has been tested on data measured with an ultra-wideband (UWB) video impulse radar (VIR) system developed by the International Research Centre for Telecommunications and Radar (IRCTR). It was found that the algorithm is able to detect all landmines (including difficult to detect M14 mines) and classifies almost all landmines correctly with a large reduction in the number of false alarms caused by the clutter. It turns out that for clutter removal it is most effective to eliminate detected objects with a small height.@en